SWE-bench Verified - <15 Min Fix
Code · 2024-08-13
The <15 min fix split of SWE-bench Verified. This child benchmark separates a source-reported subtask or subtrack from the parent aggregate so scores at different grains do not share one benchmark_slug.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Opus 5 | 98.5 |
| GPT-5.6 Sol | 97.4 |
| GPT-5.6 Luna | 96.4 |
| Claude Fable 5 | 95.9 |
| Grok 4.5 | 93.8 |
| Opus 4.8 | 92.8 |
| Kimi K3 | 92.3 |
| GPT-5.5 | 91.8 |
| Opus 4.6 | 90.2 |
| Opus 4.7 | 90.2 |
| GPT-5.3-Codex | 90.2 |
| GPT-5.2 | 89.2 |
| Gemini 3.1 Pro Preview | 88.7 |
| GLM-5.1 | 88.7 |
| Kimi K2.7 Code | 88.7 |
| Gemini 3 Pro Preview | 88.1 |
| GLM-5.2 | 88.1 |
| GPT-5.2-Codex | 88.1 |
| GPT-5.4 | 88.1 |
| Muse Spark 1.1 | 88.1 |
| Opus 4.5 | 87.6 |
| DeepSeek-V4-Pro | 87.6 |
| Gemini 3 Flash Preview | 87.6 |
| Kimi K2.6 | 87.6 |
| Sonnet 4.6 | 87.1 |
| Gemini 3.6 Flash | 87.1 |
| Grok 4.20 | 87.1 |
| MiniMax M2.5 | 87.1 |
| Composer 2.5 | 86.6 |
| Gemini 3.5 Flash | 86.6 |
| Inkling | 86.1 |
| MiniMax M2.7 | 86.1 |
| Gemini 3.5 Flash-Lite | 85.6 |
| GPT-5.4 Mini | 85.6 |
| MiniMax M2.1 | 85.6 |
| Qwen3.6-Max-Preview | 85.6 |
| Qwen3.6 Plus (2026-04-02) | 85.6 |
| Muse Spark | 85.1 |
| Kimi K2.5 | 84.5 |
| MiniMax M3 | 84.5 |